Logo

Machine Learning

Model Performance

AI Generated Summary

Churn Prediction v3.2 accuracy has drifted 2.1% over 14 days, with recall declining faster than precision. Inference latency remains healthy at 82ms. Retraining is recommended within 48 hours.

94.8% accuracy 82ms avg latency 2.1% accuracy drift · retrain recommended

Accuracy

94.8%

Precision

92.1%

Recall

90.5%

F1 Score

91.3%

ROC Analysis

Receiver operating characteristic · AUC 0.94

Evaluated on 90-day holdout set 94.8% accuracy

Model Drift Detection

Accuracy drift vs. threshold · last 5 weeks

Reviewed by Priya Patel · Data Science

Inference Latency

p50 / p95 / p99 · last 24 hours

82ms average · 18.4M predictions/day 99.98% uptime

Resource Utilization

CPU / memory / GPU during inference

Owned by Marcus Lee · Infra

Production Comparison

Current model vs. previous version vs. industry benchmark

v3.2 outperforms v3.1 baseline by 1.4 pts View version history

Performance Timeline

Drift threshold exceeded · retraining recommended

2 hours ago

Weekly benchmark evaluation completed

1 day ago

Inference latency held steady at 82ms p50

2 days ago

Tracked by Dana Ruiz · ML Performance

Optimization Recommendations

Schedule retraining tonight using the last 90 days of data to correct the 2.1% accuracy drift.
Enable batch inference for low-priority requests to reduce p95 latency by an estimated 12ms.
2 recommendations pending review

Model Health Summary

82%

Health score · Warning

Retrain recommended within 48h

Last checked 6 minutes ago Open health center